Augusto Oliveira’s Post

Lately, I tried something different. Instead of building everything the usual way, I used AI to develop a complete Angular module—from structure to state management to tests. Not perfectly. Not in one go. But enough to see a shift. It made me realize that development is slowly changing. It’s becoming less about writing every line yourself, and more about guiding, refining, and making decisions. AI didn’t replace the work—it changed the way the work happens. Still early. Still learning. But this feels like a meaningful shift in how we build software. #AI #Frontend #Angular #SoftwareEngineering #Learning

Did you know? GitHub research suggests that devs using AI assistance can complete certain tasks 55% faster! 😳 See how we use it to accelerate legacy migration at House of Angular 👇 1️⃣ Assisted Coding AI analyzes legacy code faster, suggests refactoring, and generates unit tests for existing code. 2️⃣ AI in system analysis Models can be used to map dependencies in a monolith, detect dead code, and even help document business logic. 3️⃣ AI as a migration assistant When rewriting components, AI can help generate API adapters, assist with frontend conversion, and create documentation/mappings. In short, AI is becoming a real productivity multiplier. ‼️ But great results do not happen automatically. They depend on teams that know how to verify, refine, and thoughtfully apply AI-generated output. → Without that, speed can easily come at the expense of quality. P.S. We’re working on something special for those interested in AI in software development 👀 Follow us to stay updated! #Angular #AICoding #Frontend #EnterpriseAngular

🚀 𝗜’𝗺 𝘃𝗲𝗿𝘆 𝗵𝗮𝗽𝗽𝘆 𝘁𝗼 𝗮𝗻𝗻𝗼𝘂𝗻𝗰𝗲 𝘁𝗵𝗲 𝗻𝗲𝘅𝘁 𝗰𝗼𝘂𝗿𝘀𝗲 𝗵𝗲𝗿𝗲 𝗮𝘁 𝘁𝗵𝗲 𝗔𝗻𝗴𝘂𝗹𝗮𝗿 𝗨𝗻𝗶𝘃𝗲𝗿𝘀𝗶𝘁𝘆! 🤖 𝗔𝗻𝗴𝘂𝗹𝗮𝗿 𝗔𝗜 𝗜𝗻 𝗗𝗲𝗽𝘁𝗵 𝗖𝗼𝘂𝗿𝘀𝗲 ⚡️ (𝗪𝗶𝘁𝗵 𝗖𝗹𝗮𝘂𝗱𝗲 𝗖𝗼𝗱𝗲) We are going to vibe code an Angular AI assistant chatbot from scratch using the OpenAI API and the Claude Code AI coding agent. 📚𝗔𝗯𝗼𝘂𝘁 𝘁𝗵𝗶𝘀 𝗰𝗼𝘂𝗿𝘀𝗲 Our job as developers is no longer to write most of the code ourselves. It’s now about designing systems, making architectural decisions, choosing the right dependencies, and guiding AI to generate code that is clean, maintainable, and easy to verify. The developers who know how to work effectively with AI already have a huge advantage. 🚀 This course is a complete, practical guide to AI-assisted modern Angular development with signals, focused on modern Angular with standalone components, signals, zoneless architecture, and coding agents like Claude Code. We start by configuring Claude Code properly for Angular, Node, and TypeScript projects, including the Angular MCP Server, so the AI has access to up-to-date Angular documentation and best practices for your exact Angular version. 🧠 Then we use it to build a complete full-stack application from scratch: ✨ A signals-based Angular AI chat interface ⛭ Backed by a Node + Express REST API 🤝 Integrated with the OpenAI API 🔐 Authentication, JWT sessions, route protection, and secure password handling This includes generating the frontend UI from Figma screenshots, building the full chat experience, and implementing the backend architecture step by step. Most importantly, you’ll learn how to actually work with AI professionally: 👉 What to delegate 👉 What to review 👉 How to structure prompts 👉 How to stop the AI from repeating the same mistakes This is a practical, code-first course built around one real sample application. ⏳ The course will be pre-launched in the coming days. 𝗛𝗶𝘁 𝗿𝗲𝗽𝗹𝘆 — what do you think about this course? 👇 🔥 #Angular #AI #ClaudeCode #OpenAI #TypeScript #WebDevelopment #SoftwareEngineering

🚀 AI in VS Code for Angular Developers – Series 🚀 Day 10 – Ultimate AI + Angular Cheat Sheet (Save This) You’ve seen everything in this series. Now here’s: 👉 One post to remember it all 🔥 The Reality AI can: ✔ Generate code ✔ Fix errors ✔ Write tests ✔ Optimize performance But only if: 👉 YOU use it correctly 🔹 Ultimate Cheat Sheet ⚡ Code Generation 👉 Use Copilot / Cursor 👉 Generate components, services ⚡ Debugging 👉 Ask AI to explain errors 👉 Identify root cause ⚡ Testing 👉 Generate Jasmine/Karma tests 👉 Review before committing ⚡ Refactoring 👉 Remove duplication 👉 Improve structure ⚡ Performance 👉 Optimize change detection 👉 Reduce unnecessary renders ⚡ Security 👉 Follow OWASP practices 👉 Validate inputs 🔹 Best Tools (Quick Guide) ✔ Copilot → Fast inline coding ✔ Cursor → Deep understanding ✔ Codeium → Free alternative 🎯 Golden Rules ✔ Guide AI with clear prompts ✔ Always review output ✔ Never blindly trust 🧠 Final Insight 👉 AI will not replace developers. 👉 But developers using AI will replace those who don’t. 💬 What’s your biggest learning from this series? 🚀 Follow for Next Series – (We go even deeper 🚀) #Angular #VSCode #AIDevelopment #DeveloperProductivity #FrontendEngineering

Most developers use AI to write code. I started using it to think better. For the past 1 month, I’ve been using Claude AI alongside my full-stack work (Java + Spring Boot + Angular), especially while building real-time systems using WebRTC and AI integrations. Here’s what actually changed for me: 🔹 Designing APIs before writing them Instead of jumping into controllers, I started breaking down: • Request/response flow • Validation layers • Edge cases Example: • What if input is partial? • Where should validation live? • How should errors be handled consistently? This reduced rework significantly. 🔹 Debugging async issues (Angular + APIs) With RxJS and chained API calls, bugs aren’t obvious. I started analyzing: • API responses • Observable flows • Error logs and asking: “Where can this break logically?” Much faster than trial-and-error debugging. 🔹 Real-time systems (WebRTC + chat/audio) While building real-time communication features, I used it to reason about: • Offer/answer lifecycle • Failure scenarios (reconnect, drops) • State sync between frontend and backend This felt like having a second reviewer. 🔹 Frontend performance thinking While optimizing Angular apps: • Lazy loading • Component design • State management I stopped blindly applying patterns and started asking: “Is this optimal for my use case?” 🔹 The biggest shift: I moved from: 👉 “How do I write this code?” to: 👉 “Is this the right way to design this system?” One realization: AI doesn’t make you a better developer. It exposes your level. If fundamentals are weak → shallow answers If fundamentals are strong → powerful thinking tool Currently exploring this deeper in: • Microservices design • API contracts • Real-time + AI system architecture Curious how others are using AI in real engineering workflows. #FullStack #Java #SpringBoot #Angular #ClaudeAI #WebRTC #Microservices #SoftwareEngineering #Electron #Node #LLM #RAG

"Why does my AI keep writing outdated Angular code?" Because it doesn't know about Angular v21. The fix? One command: $ ng mcp Angular's MCP Server gives your AI real-time access to: ✓ Official best practices ✓ Curated code examples ✓ Your project structure ✓ Angular documentation ✓ Migration tools ✓ An AI Tutor 30 seconds to set up. Works with VS Code, Cursor, JetBrains, Claude Code. I made a visual guide breaking down every tool. Swipe through ↓ Your AI just leveled up. #Angular #MCP #AI #WebDevelopment #Frontend

Today I built a Next.js TypeScript application for a customer, and it reinforced something I keep coming back to: no matter how much AI advances, the fundamentals of software engineering do not change. Multi-agent AI can accelerate output. It does not guarantee correctness. Research shows roughly 45% of AI-generated code contains security flaws while appearing production-ready . “Vibe coding” where you generate code without understanding it is just skipping fundamentals with extra steps, and the result is always the same: bugs, debt, and fragility. What actually worked was going back to TDD. Tests define the behavior, drive the design, and act as the only reliable control mechanism over AI-generated code. And before someone points it out, this is not about chasing 100% coverage. Any seasoned practitioner knows that was never the goal, and it was not here either. The goal was clarity of intent. I knew what I wanted built and knew how to guide AI to help produce the tests, which just happened to ensure my requirements were met. AI does not replace discipline, it raises the bar for it. Taking shortcuts to “engineer” something is still failure, just faster now. AI has not changed that. It has only made it more obvious.

When most people say AI is going to replace developers, they’re usually thinking about the top half of this image. If "software engineering" was just about writing basic HTML, CSS, and a few lines of JavaScript to make a button click, then sure, the robots would have won a long time ago. 😂 But the reality of modern development is the bottom half of this image. The job has evolved far beyond "writing code." In 2026, being an engineer means navigating an absolute ocean of complexity. It’s not just about the syntax; it’s about: Architecture & State Management: Choosing between React, Vue, or Angular, and managing data with GraphQL or Redux. Infrastructure & DevOps: Orchestrating containers with Docker and Kubernetes, and managing cloud scale on AWS, Azure, or GCP. Data Strategy: Deciding when to use a relational DB like Postgres versus a NoSQL powerhouse like MongoDB or Redis. The Ecosystem: Dealing with build tools like Webpack, transpilers like Babel, and the type-safety of TypeScript. The Truth About AI in Engineering: AI is a tool, not a replacement. It’s great at the "then", the repetitive, boilerplate syntax. But it struggles with the "now", the high-level decision-making, the complex integration of fragmented systems, and the problem-solving required to keep these massive stacks running. AI won't replace developers, but it might replace people who only know how to write code. Real software engineering is about system design, logic, and managing complexity. The stack is bigger than ever, the stakes are higher, and the need for human engineers who can navigate this chaos has never been greater. #SoftwareEngineering #WebDevelopment #AI #TechTrends #FullStack #CodingLife #FutureOfWork

Ryan Atkinson published an interesting study counting LLM tokens across 18 JS frameworks for the same UI patterns. The results: Marko uses 22% fewer tokens than Svelte. Angular uses 101% more. Sounds like Angular is the worst choice for AI coding, right? Not so fast. I dug into the actual code snippets and found that frameworks using fewer tokens tend to be more implicit. They hide lifecycle hooks, component names, and type information. That's exactly what makes AI hallucinate more. Explicit code gives LLMs better context, stronger type checking, and more reliable AST analysis. A framework that costs 3x more tokens but generates correct, strongly-typed, debuggable code is cheaper than one that needs 10 re-runs. Link to full article is in the comment. #programming #AI #frontend #webdev #angular

🚀 AI in VS Code for Angular Developers – Series 🚀 Day 8 – Common AI Mistakes Developers Make (Avoid These) AI is powerful. But most developers are using it wrong. 🔥 The Reality AI is NOT: ❌ Always correct ❌ Always optimized ❌ A replacement for thinking 👉 It’s just a tool. 🔹 Top Mistakes Developers Make ❌ Blindly Accepting AI Code 👉 “It works” ≠ “It’s correct” ✔ Always validate ✔ Understand before using ❌ Skipping Important Setup 👉 AI may generate code but miss: • Imports • Providers • Edge cases ✔ Always review generated code ❌ Over-Reliance on AI Chat 👉 Constantly asking AI = slow ✔ Use inline suggestions first ✔ Use chat only when needed ❌ Ignoring Best Practices 👉 AI may not follow: • Angular architecture • Clean code principles ✔ Guide AI with better prompts 🔹 Which Tools Help Avoid This? Inside VS Code: 🟢 GitHub Copilot 👉 Fast inline suggestions 👉 Less dependency on chat 🟣 Cursor 👉 Better context understanding 👉 Good for deeper analysis 🔵 Codeium 👉 Free option 👉 Good for quick suggestions 🎯 Pro Tip (High Value 🔥) Instead of: ❌ “Write code” Ask: Write Angular code following best practices, proper structure, and performance optimization 👉 You guide AI → better output 🧠 Enterprise Insight 👉 Good developers use AI. 👉 Great developers: 👉 Control AI ⚠️ Final Rule ✔ Trust AI → partially ✔ Trust your knowledge → fully 💬 Have you ever faced issues due to AI-generated code? 🚀 Follow for Day 9 – AI + Secure Coding (OWASP for Frontend) #Angular #VSCode #AIDevelopment #DeveloperProductivity #FrontendEngineering